Due to long-term operation in humid, high-temperature, and dusty underground environments, coal mine belt conveyors are prone to various malfunctions, such as belt deviation, slippage, damage, breakage, as well as motor and roller failures, which require the equipment to be shut down for maintenance, resulting in increased coal production costs and decreased production efficiency. Based on the large-scale AI architecture featuring cloud-edge collaboration and on-the-fly edge learning, this paper integrates computer vision, equipment control, and big data analytics technologies to propose an intelligent monitoring algorithm for coal flow transportation systems. It develops key AI model technologies for conveyor belt anomaly detection, ultimately establishing a visual intelligent monitoring and management system for coal transportation. The implementation achieves intelligent speed regulation of conveyor belts based on coal volume monitoring, while forming a closed-loop operational process that encompasses damage tracking, anomaly warning, and incident handling throughout the conveyor belt operation. The application has been successful in Hetaoyu coal mine, belt speed measurement error <±2%, coal volume statistics error < ±5%, 30% reduction in belt idling rate, annual electricity savings of approximately 150,000 kWh per kilometer of conveyor belt, reduced mechanical impact intensity, 40% extension of roller replacement cycle, reducing the number of belt inspection workers for 6 km of the main inclined shaft by 10 times per day, reducing abnormal downtime by 15 minutes per day, extending the service life of the belt, significantly reducing staff and increasing efficiency, and creating a direct economic benefit of RMB 2 million per year and an indirect economic benefits exceeding RMB 10 million per year. The application effect is good.